Instagram ads for SaaS companies

Start with what the Browse page shows rather than with a claim. Every one of the ten SaaS ads opens on a moment and not a feature: a Sunday night rota that takes three hours, a phone turned face-down because the pager is somebody else's problem tonight, an empty notepad after a client call, a spreadsheet quietly winning an argument it should have lost. That is not a stylistic preference, it is the platform. Instagram is where people are between things, and a person between things will recognise a Sunday night before they will recognise a product category. On the Facebook twin of this page — feed, right column, search — you have surfaces where somebody is arguably looking for something. Here you have surfaces where they are not, and the first line of the ad has to do the work that intent would otherwise have done.

The second Instagram-specific difference is who is on it. Pew's November 2025 survey puts Instagram at 80% of adults aged 18 to 29 and half of all US adults, while Facebook reaches 80% of 30- to 49-year-olds. In software that maps onto a distinction worth planning around: the practitioner who will use the tool daily skews to the first group, and the manager who approves the invoice skews to the second. That is a heuristic rather than a fact about your funnel, but it is a useful one — if you sell product-led, where the user and the buyer are the same person, Instagram is where that person is and the split collapses in your favour. If you sell to a committee, expect this app to fill the top of the funnel and the other one to close it.

Third, shape, and this is the biggest practical gap between the two cells of this cluster. Instagram gives you 4:5 for feed and full-screen 9:16 for Stories and Reels — eight of the ten ads on the Browse page are 4:5 and two are 9:16 — and it gives a 1.91:1 asset nowhere useful to go, because our publisher routes that ratio to Facebook's right-hand column and search results and to no Instagram position. (Instagram Feed's own technical requirements do accept up to 1.91:1 — corrected 7 September 2026, this paragraph previously said the ratio did not exist here. It exists; the surfaces built for it are Facebook's.) For most industries that costs you a placement. For software it costs you the comparison ad: "X vs Y" is where a large share of SaaS budget goes, and it works best against somebody who has just typed a competitor's name. There is no search placement here, so a comparison has to win in-feed against a reader who was not asking — which is why the contrarian ad on the Browse page argues a position ("47 charts or one answer") rather than naming a rival.

Finally, the screenshot problem, which is specific to this industry and worse on this app than anywhere else. The instinct is to show the product. A dashboard rendered at 1080 × 1350 and viewed on a phone at arm's length is an illegible grey rectangle, and a 9:16 story is worse. Six of the ten ads on the Browse page render a situation instead of a UI — a checklist, a notepad, a whiteboard being retired — and the two that do show product detail show one number from it rather than the screen it lives on.

This is the one section of this cluster where the recommendation inverts. Every other industry on this site is Leads-dominated because nothing they sell can complete inside the ad's own session — nobody books an implant or buys a house from a feed. Software can. A free trial starts in a browser tab and a self-serve subscription is a card form away, which puts two objectives on the table that a dental practice cannot use and takes the pressure off the one it depends on.

The evidence is in the batch. The ten ads on the Browse page were pinned as 5 Traffic, 3 Leads, 2 Sales and no Awareness — the only page on this site where Traffic leads, and the only one where Sales appears at all. The reason Traffic leads is not timidity: an ad cannot explain software. What it can do is make one person recognise one problem and then get out of the way, and the page you send them to does the selling. Buying Leads for a product that is trying to be self-serve is how you end up with an email address belonging to somebody who has never seen the thing.

Sales is the right objective once the funnel can actually feed it — meaning the pixel or Conversions API is firing a real Subscribe or StartTrial event and doing so often enough to optimise on. Below that volume it is a worse Traffic campaign. The budget section has the arithmetic for deciding which side of that line you are on.

Software has no offer list in the sense a dentist does — there is one product. What varies is the argument, and the unit of planning here is a role plus a moment: which person, in which recurring situation, recognises themselves. Every row below is one of those pairs, taken from the audience briefs that actually produced the Browse page's ads.

Each pins four decisions: who it is for, the angle family that argues it (each is a page in the angle library), the objective to buy it under, and the finished ad that shows it assembled. Two rows have no example ad, and the reason is worth knowing rather than hiding: the SaaS batch used only six of the ten angle families, and the four it skipped — Offer & Economics, Product & Mechanism, Scarcity and Founder POV — are the four that are hardest to do honestly for software. Pricing arguments need a real price, mechanism arguments need a mechanism a reader can picture, and manufactured urgency in a category with a free tier reads as a lie. They are still worth running; they are just the rows you have to write yourself.

One structural note that holds across all of them. The first line names the role, because the ad set cannot. "Independent restaurant GMs" is a superb targeting instrument and it is not available in the picker, so it has to be inside the picture.

First, the good news, and it is worth stating because two of the three guides in this cluster open with the opposite. Software is not a special ad category — Meta accepts exactly four, for housing, employment, financial products and services, and social issues, elections and politics — so nothing here is taken away from you. Age is yours, geography is yours, lookalikes are yours, exclusions are yours. There is no policy wall around the copy either.

The bad news is that none of those levers asks the question you actually have. You want people who run a restaurant, or who carry a pager, or who own a small agency. Meta's detailed targeting has been narrowed repeatedly over the last few years and the work-related categories are the ones most often affected, so the honest instruction is the same one the real-estate guide gives about custom audiences: open the picker in your own account, see what is genuinely in it today, and build the plan from that rather than from any guide's description of it, this one included. It costs you nothing to check and it is the only way to be right about it in six months.

In practice the recommendation does not change either way, which is why the uncertainty above is tolerable: go broad and let the creative filter. A first line that says "independent restaurant GMs" — or renders a Sunday night rota that only a restaurant GM has ever stared at — excludes everybody else more reliably than an interest category built from what people clicked three years ago, and it does it without shrinking the audience below the size the delivery system needs. Every audience brief on the Browse page is a job description and none of them was ever selectable.

Where the ad set does earn its keep is the second stage. Website custom audiences, video viewers and engagement audiences are all available here, and lookalikes are too — which is a real advantage over the housing cluster and worth using properly. Seed lookalikes from paying customers, not from trial signups: a free-trial list in a product-led company is largely people who were curious, and modelling an audience on curiosity is how a campaign scales the wrong half of the funnel. Geography deserves one deliberate decision rather than a default: for a global product the temptation is worldwide, and the cost is a stream of cheap impressions in markets you cannot support, price for, or invoice.

A landing page, almost always — which is the exact opposite of the dentist guide's recommendation, and the reason is the product rather than the platform. Software has to be seen before anyone can want it. An Instant Form is superb at converting a person who has already decided and needs only to hand over a phone number; it is useless when the thing you are asking them to decide about is a screen they have never looked at. Collect an email from somebody who has never seen the product and you have bought a name and a sequence of emails to send it.

So send the click to a page that lets them start. Not the homepage, which answers a different question than the ad asked, and — this one is common — not the pricing page. Pricing is where a person goes once they want the thing; putting it in front of a reader who is still deciding whether the Sunday-night rota problem is solvable asks them to evaluate cost before value. Match the page to the argument the ad made, and let the primary action on it be the trial, the sandbox, or whatever the smallest real use of the product is.

The exception is a genuine sales-led motion. If the product has no self-serve path, if the tier being advertised is quoted rather than priced, or if implementation requires a conversation before anyone can use anything, then a booked demo is the conversion and an Instant Form on its higher-intent type — which adds a review step — with two or three qualifying questions is the right instrument. That is a different business from the one most of this page is written for, and running its playbook on a product-led product is the most common way a SaaS Meta account underperforms.

Whichever you pick, instrument it before you spend. The destination decision and the optimisation-event decision are the same decision made twice, and the budget section below is about the second half of it.

The single most useful rule in this industry is: do not put the product in the ad. Not because the product is uninteresting, but because a dashboard rendered at 1080 × 1350 and looked at on a phone, at arm's length, in a feed, is a grey rectangle with unreadable text in it. Six of the ten ads on the Browse page render a situation rather than a screen — a checklist, a whiteboard, an empty notepad, a phone face-down — and the ones that do use product detail lift one number out of the interface rather than photographing the interface.

The second rule is that the first line is your targeting, and it has a harder job here than on the other two pages in this cluster. A dentist's reader has a toothache; a seller's listing has expired. Your reader is having an ordinary evening and has not thought about their rota since Friday. So open on the recurring moment rather than the problem category, and be specific enough to be exclusionary — three hours every Sunday is recognisable, "streamline your scheduling" is not.

Third, one number, and one you actually measured. This category is saturated with unfalsifiable adjectives, which is exactly why a figure with a window attached — 90% accuracy by week two, four minutes per invoice — reads as a different kind of claim. It is also the reason the invented metrics on the Browse page are labelled as placeholders rather than quietly left to look real: the format only works while the number is true.

Three stages, and unlike the other two clusters the third one is a revenue stage rather than a courtesy. Software has expansion and churn, so the audience you already own is worth advertising to on its own terms — not with the acquisition ad, which they will find irritating, but with the feature they are not using and the plan they have outgrown.

The split is deliberately less cold-heavy than a local business needs, because the second stage is where a software funnel actually converts: cold buys recognition, warm buys the trial, and the ad set only becomes precise once there is behaviour to target.

Here is the one place on this site where the learning-phase arithmetic works in your favour, and it is worth doing before anything else. Meta's delivery system needs roughly 50 optimisation events per ad set per week to leave the learning phase; below that an ad set sits in "learning limited", which is an admission that it cannot optimise with what it has been given. For a dental practice, 50 booked implant consultations a week is not a real number. For a self-serve software product, 50 trial starts a week is entirely achievable — which means you can actually buy the optimisation everyone else on this site is priced out of.

The trap is right behind it. Meta's standard events include StartTrial ("when a person starts a free trial"), Subscribe ("when a person applies to start a paid subscription"), CompleteRegistration and Lead, and the easy choice is to optimise for the cheapest of them. Optimise for StartTrial and the system will faithfully find you the people most likely to start a trial, which is not the same population as the people most likely to pay — and after a month you will have a beautiful cost per trial and a conversion rate that has quietly halved. So: use StartTrial while you are below the volume threshold, and move to Subscribe, or to a qualified-activation event you define, as soon as that event clears roughly 50 a week. Sending the deeper event server-side through the Conversions API is usually what makes that possible, because the moment that matters often happens after the browser has gone.

Then leave it alone. Pausing an ad set, or changing its optimisation event, audience or creative, is a significant edit and restarts the learning phase; budget and bid changes can too, depending on size. Changing the optimisation event is the biggest one available to you here, so make that switch deliberately and once, not as a weekly experiment.

Judge on payback rather than on cost per acquisition. It is the number this industry has that the others do not: a signup is worth a monthly figure over a lifetime rather than a one-off, so the honest question is how many months of subscription it takes to repay the ad that produced it, and whether your company can wait that long. Seasonality is mild compared with the other clusters — the real calendar effects are the ones inside your own business, like a pricing change or a launch, and those are the moments worth concentrating spend on.

The first two are the reason most SaaS Meta accounts underperform, and they are the same mistake at two different layers: treating the ad as a place to explain the product.